Hi all, Two questions: 1. What is the recommended approach for accessing data from another account with data in an s3 bucket that can only be pulled in through an assumed role. I saw an

@assume_role

decorator that users could potentially leverage? 2. Relating to GDPR compliance, we might have to ensure that metadata in the s3 buckets are deleted after a few days. We only want to delete the snapshot data/data artifacts though, not the data related to the flow itself that is used to populate the UI. There are two different data folders in s3:

data/

and

metaflow/FlowName/data

. What is the difference?

1. Yeah, you can assume role in a task and instantiate S3 with the new role. We don’t have baked-in support for it yet but it should work https://outerboundsco.slack.com/archives/C02116BBNTU/p1645168826272899?thread_ts=1645165273.396909&channel=C02116BBNTU&message_ts=1645168826.272899

Got it - sounds good. Thanks!

flowname/data

contains artifacts,

data/

is the default location for

metaflow.S3(run=self)

Perfect! Thanks Ville.